文献类型：期刊文献

英文题名：Shelf Product Detection Based on Deep Neural Network

作者：Geng, Zelong[1]; Wang, Zhongze[1]; Weng, Tiandong[1]; Huang, Yuhui[1]; Zhu, Yu[1]

机构：[1] School of Information Science and Engineering, East China University of Science and Technology, Shanghai, China

年份：2019

外文期刊名：Proceedings - 2019 12th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics, CISP-BMEI 2019

收录：EI(收录号：20200708158522)

语种：英文

外文关键词：Statistical tests - Multilayer neural networks - Image segmentation - Edge detection - Object detection

摘要：Nowadays, there is a high demand for product detection in the background of shelves. Considering the structure of the shelf and the placement of products, this paper proposes a shelf product detection method based on RFBNet and combined with traditional image processing methods. The method firstly uses the edge detection and other image preprocessing methods to separate the shelves layer by layer and then detects the image of each shelf layer by the deep neural network. Finally, eliminate incorrect results based on the placement characteristics of the product, thereby improving the detection accuracy. We established a synthetic training dataset by a randomly pasting method, and the images in test dataset are captured in real scenes. The results show that this method can obtain better detection accuracy on test images with 0.95 for mAP and 0.89 for F1-Score. ? 2019 IEEE.